That assumes that you should just expose your Postgres tables as your data. As someone who consumes more APIs than they write nowadays, I appreciate when data

That assumes that you should just expose your Postgres tables as your data. As someone who consumes more APIs than they write nowadays, I appreciate when data is scoped to task[0]. Employing window functions and stored procedures that can make queries into data fit for purpose is the ideal - databases are faster at manipulating data than any intermediate language, most of the time. Unfortunately, I don't see them employed enough. Backend developers seem content with throwing the schema over the wall way too often. [0]: As an aside, this is why I personally like GraphQL so much. The middleware lets me cleanup data that backend engineers simply refuse to do, most of the time.